Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Found mnist Data
Found celeba Data

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x7f4c38900588>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7f4c387e34e0>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.1.0
Default GPU Device: /gpu:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    # TODO: Implement Function
    inputs_real = tf.placeholder(tf.float32, (None, image_width, image_height, image_channels), name='input_real')    
    inputs_z = tf.placeholder(tf.float32, (None, z_dim), name='input_z')
    learning_rate = tf.placeholder(tf.float32, name="learning_rate")

    return inputs_real, inputs_z, learning_rate


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

def discriminator(images, reuse=False):
    """
    Create the discriminator network
    :param images: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    alpha = 0.2
    # TODO: Implement Function
    with tf.variable_scope("discriminator", reuse = reuse):      
        # Input layer is 28x28x3        
        x1 = tf.layers.conv2d(images, 64, 5, strides=2, padding='same')
        relu1 = tf.maximum(alpha * x1, x1)
        # 16x16x64
        
        x2 = tf.layers.conv2d(relu1, 128, 5, strides=2, padding='same')
        bn2 = tf.layers.batch_normalization(x2, training=True)
        relu2 = tf.maximum(alpha * bn2, bn2)
        # 8x8x128
        
        x3 = tf.layers.conv2d(relu2, 256, 5, strides=2, padding='same')
        bn3 = tf.layers.batch_normalization(x3, training=True)
        relu3 = tf.maximum(alpha * bn3, bn3)
        # 4x4x256

        # Flatten
        flat = tf.reshape(relu3, (-1, 4*4*256))
        logits = tf.layers.dense(flat, 1)
        out = tf.sigmoid(logits)
        
    return out, logits


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z, out_channel_dim, is_train=True):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
    # TODO: Implement Function
    alpha=0.1
    with tf.variable_scope("generator", reuse= not(is_train)):
        #Fully connected layer
        x1 = tf.layers.dense(z, 7 * 7 * 256, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))

        #Reshape the fully connected layer to start the convolutional stack
        x1 = tf.reshape(x1, (-1, 7, 7, 256))
        x1 = tf.layers.batch_normalization(x1, training=is_train)
        x1 = tf.maximum(alpha * x1, x1)
        # 7x7x512 now
        
        x2 = tf.layers.conv2d_transpose(x1, 256, 5, strides=2, padding='same')
        x2 = tf.layers.batch_normalization(x2, training=is_train)
        x2 = tf.maximum(alpha * x2, x2)
        # 14x14x256 now
        
        x3 = tf.layers.conv2d_transpose(x2, 128, 5, strides=2, padding='same')
        x3 = tf.layers.batch_normalization(x3, training=is_train)
        x3 = tf.maximum(alpha * x3, x3)
        # 28x28x128 now
        
        # Output layer
        logits = tf.layers.conv2d_transpose(x3, out_channel_dim, 5, strides=1, padding='same')
        # 28x28x3 now
        
        out = tf.tanh(logits)   
        
    return out


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    # TODO: Implement Function
    #using the parameter smooth. Typically used with classifiers to improve performance.
    smooth = 0.1
    g_model = generator(input_z, out_channel_dim)
    d_model_real, d_logits_real = discriminator(input_real, reuse=False)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse = True)
    
    
    d_loss_real = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits = d_logits_real,labels = tf.ones_like(d_model_real) * (1 - smooth)))
    d_loss_fake = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.zeros_like(d_model_fake)))
    g_loss = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.ones_like(d_model_fake)))

    d_loss = d_loss_real + d_loss_fake
    
    return d_loss, g_loss


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    # TODO: Implement Function
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]
    
    #Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1 = beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1 = beta1).minimize(g_loss, var_list=g_vars)
        
    return d_train_opt, g_train_opt


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    # TODO: Build Model
    steps = 0
    print_every = 10
    show_every = 100
    
    _, image_width, image_height, image_channels = data_shape
    input_real, input_z, lr = model_inputs(image_width, image_height,image_channels, z_dim)
    d_loss, g_loss = model_loss(input_real, input_z,image_channels)
    d_opt, g_opt = model_opt(d_loss, g_loss, lr, beta1)
    
    is_train = tf.placeholder(tf.bool)  
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        for epoch_i in range(epoch_count):
            for batch_images in get_batches(batch_size):
                # TODO: Train Model
                steps += 1
                batch_images = batch_images * 2
                
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))
                # Run optimizers
                _ = sess.run(d_opt, feed_dict={input_real: batch_images, input_z: batch_z, lr:learning_rate})
                _ = sess.run(g_opt, feed_dict={input_real: batch_images, input_z: batch_z, lr:learning_rate})
                               
                if steps % print_every == 0:
                    # At the end of each epoch, get the losses and print them out
                    train_loss_d = d_loss.eval({input_z: batch_z, input_real: batch_images})
                    train_loss_g = g_loss.eval({input_z: batch_z})

                    print("Epoch {}/{} Step {}...".format(epoch_i+1, epoch_count, steps),
                      "Discriminator Loss: {:.4f}...".format(train_loss_d),
                      "Generator Loss: {:.4f}".format(train_loss_g))
    
                if steps % show_every == 0:
                    #Print the current image
                    show_generator_output(sess, 25, input_z, image_channels, data_image_mode)

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 64
z_dim = 100
learning_rate = 0.0002
beta1 = 0.5


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch 1/2 Step 10... Discriminator Loss: 0.4162... Generator Loss: 4.6087
Epoch 1/2 Step 20... Discriminator Loss: 0.4501... Generator Loss: 2.5648
Epoch 1/2 Step 30... Discriminator Loss: 0.4473... Generator Loss: 5.5668
Epoch 1/2 Step 40... Discriminator Loss: 0.8144... Generator Loss: 1.7805
Epoch 1/2 Step 50... Discriminator Loss: 0.5995... Generator Loss: 2.1799
Epoch 1/2 Step 60... Discriminator Loss: 0.9210... Generator Loss: 1.9464
Epoch 1/2 Step 70... Discriminator Loss: 0.7397... Generator Loss: 1.8119
Epoch 1/2 Step 80... Discriminator Loss: 1.0817... Generator Loss: 2.0575
Epoch 1/2 Step 90... Discriminator Loss: 1.3563... Generator Loss: 2.5081
Epoch 1/2 Step 100... Discriminator Loss: 1.1096... Generator Loss: 0.9838

Epoch 1/2 Step 110... Discriminator Loss: 1.1064... Generator Loss: 0.8206
Epoch 1/2 Step 120... Discriminator Loss: 1.1948... Generator Loss: 0.9123
Epoch 1/2 Step 130... Discriminator Loss: 1.5127... Generator Loss: 0.4818
Epoch 1/2 Step 140... Discriminator Loss: 1.0324... Generator Loss: 1.4577
Epoch 1/2 Step 150... Discriminator Loss: 1.0713... Generator Loss: 1.0183
Epoch 1/2 Step 160... Discriminator Loss: 1.0482... Generator Loss: 1.7654
Epoch 1/2 Step 170... Discriminator Loss: 1.1345... Generator Loss: 1.9795
Epoch 1/2 Step 180... Discriminator Loss: 0.9310... Generator Loss: 1.0294
Epoch 1/2 Step 190... Discriminator Loss: 1.1152... Generator Loss: 0.7662
Epoch 1/2 Step 200... Discriminator Loss: 1.1132... Generator Loss: 1.6096

Epoch 1/2 Step 210... Discriminator Loss: 0.9283... Generator Loss: 0.9879
Epoch 1/2 Step 220... Discriminator Loss: 0.9135... Generator Loss: 1.1453
Epoch 1/2 Step 230... Discriminator Loss: 0.8613... Generator Loss: 1.2542
Epoch 1/2 Step 240... Discriminator Loss: 0.9460... Generator Loss: 1.3988
Epoch 1/2 Step 250... Discriminator Loss: 0.8946... Generator Loss: 1.1722
Epoch 1/2 Step 260... Discriminator Loss: 0.8205... Generator Loss: 1.3204
Epoch 1/2 Step 270... Discriminator Loss: 0.8576... Generator Loss: 1.3610
Epoch 1/2 Step 280... Discriminator Loss: 0.9154... Generator Loss: 1.4812
Epoch 1/2 Step 290... Discriminator Loss: 1.2526... Generator Loss: 2.5330
Epoch 1/2 Step 300... Discriminator Loss: 0.8713... Generator Loss: 1.2218

Epoch 1/2 Step 310... Discriminator Loss: 0.8615... Generator Loss: 1.3869
Epoch 1/2 Step 320... Discriminator Loss: 1.0955... Generator Loss: 0.7988
Epoch 1/2 Step 330... Discriminator Loss: 1.1397... Generator Loss: 0.7575
Epoch 1/2 Step 340... Discriminator Loss: 1.0353... Generator Loss: 2.0635
Epoch 1/2 Step 350... Discriminator Loss: 1.2707... Generator Loss: 2.8639
Epoch 1/2 Step 360... Discriminator Loss: 0.9157... Generator Loss: 1.2263
Epoch 1/2 Step 370... Discriminator Loss: 1.0543... Generator Loss: 1.1553
Epoch 1/2 Step 380... Discriminator Loss: 1.0711... Generator Loss: 1.4519
Epoch 1/2 Step 390... Discriminator Loss: 1.0245... Generator Loss: 1.0673
Epoch 1/2 Step 400... Discriminator Loss: 1.1086... Generator Loss: 1.0315

Epoch 1/2 Step 410... Discriminator Loss: 1.2736... Generator Loss: 0.5852
Epoch 1/2 Step 420... Discriminator Loss: 1.1110... Generator Loss: 1.2544
Epoch 1/2 Step 430... Discriminator Loss: 1.0199... Generator Loss: 0.9989
Epoch 1/2 Step 440... Discriminator Loss: 1.0474... Generator Loss: 1.2390
Epoch 1/2 Step 450... Discriminator Loss: 0.9688... Generator Loss: 1.4482
Epoch 1/2 Step 460... Discriminator Loss: 1.1181... Generator Loss: 0.7780
Epoch 1/2 Step 470... Discriminator Loss: 0.9287... Generator Loss: 1.4584
Epoch 1/2 Step 480... Discriminator Loss: 1.0113... Generator Loss: 1.2287
Epoch 1/2 Step 490... Discriminator Loss: 1.0633... Generator Loss: 0.9347
Epoch 1/2 Step 500... Discriminator Loss: 1.0378... Generator Loss: 1.0422

Epoch 1/2 Step 510... Discriminator Loss: 1.0329... Generator Loss: 1.3286
Epoch 1/2 Step 520... Discriminator Loss: 0.9473... Generator Loss: 1.0131
Epoch 1/2 Step 530... Discriminator Loss: 0.9963... Generator Loss: 1.5298
Epoch 1/2 Step 540... Discriminator Loss: 0.9847... Generator Loss: 1.5873
Epoch 1/2 Step 550... Discriminator Loss: 1.0421... Generator Loss: 0.9462
Epoch 1/2 Step 560... Discriminator Loss: 1.0019... Generator Loss: 1.0953
Epoch 1/2 Step 570... Discriminator Loss: 0.9065... Generator Loss: 1.4824
Epoch 1/2 Step 580... Discriminator Loss: 1.1857... Generator Loss: 0.6862
Epoch 1/2 Step 590... Discriminator Loss: 1.0023... Generator Loss: 1.0988
Epoch 1/2 Step 600... Discriminator Loss: 0.9837... Generator Loss: 1.2255

Epoch 1/2 Step 610... Discriminator Loss: 1.1262... Generator Loss: 0.7545
Epoch 1/2 Step 620... Discriminator Loss: 0.9845... Generator Loss: 1.2048
Epoch 1/2 Step 630... Discriminator Loss: 0.9523... Generator Loss: 1.8169
Epoch 1/2 Step 640... Discriminator Loss: 1.3460... Generator Loss: 2.2265
Epoch 1/2 Step 650... Discriminator Loss: 1.0349... Generator Loss: 0.9005
Epoch 1/2 Step 660... Discriminator Loss: 1.0894... Generator Loss: 0.8867
Epoch 1/2 Step 670... Discriminator Loss: 1.0157... Generator Loss: 1.0711
Epoch 1/2 Step 680... Discriminator Loss: 1.0939... Generator Loss: 1.8559
Epoch 1/2 Step 690... Discriminator Loss: 1.0349... Generator Loss: 1.4791
Epoch 1/2 Step 700... Discriminator Loss: 1.0698... Generator Loss: 1.3614

Epoch 1/2 Step 710... Discriminator Loss: 1.0955... Generator Loss: 0.7910
Epoch 1/2 Step 720... Discriminator Loss: 1.0201... Generator Loss: 1.0462
Epoch 1/2 Step 730... Discriminator Loss: 1.0944... Generator Loss: 1.5817
Epoch 1/2 Step 740... Discriminator Loss: 0.9611... Generator Loss: 1.0950
Epoch 1/2 Step 750... Discriminator Loss: 1.1010... Generator Loss: 0.7973
Epoch 1/2 Step 760... Discriminator Loss: 0.9999... Generator Loss: 1.0005
Epoch 1/2 Step 770... Discriminator Loss: 1.1852... Generator Loss: 0.6859
Epoch 1/2 Step 780... Discriminator Loss: 1.0418... Generator Loss: 1.1899
Epoch 1/2 Step 790... Discriminator Loss: 1.0319... Generator Loss: 0.8962
Epoch 1/2 Step 800... Discriminator Loss: 0.9108... Generator Loss: 1.0475

Epoch 1/2 Step 810... Discriminator Loss: 1.1141... Generator Loss: 2.3206
Epoch 1/2 Step 820... Discriminator Loss: 0.9273... Generator Loss: 1.1011
Epoch 1/2 Step 830... Discriminator Loss: 1.1662... Generator Loss: 0.6753
Epoch 1/2 Step 840... Discriminator Loss: 0.8071... Generator Loss: 1.3748
Epoch 1/2 Step 850... Discriminator Loss: 1.0708... Generator Loss: 0.7503
Epoch 1/2 Step 860... Discriminator Loss: 0.9500... Generator Loss: 1.0558
Epoch 1/2 Step 870... Discriminator Loss: 0.9473... Generator Loss: 2.0067
Epoch 1/2 Step 880... Discriminator Loss: 0.8964... Generator Loss: 1.3017
Epoch 1/2 Step 890... Discriminator Loss: 1.0043... Generator Loss: 1.0989
Epoch 1/2 Step 900... Discriminator Loss: 1.0849... Generator Loss: 0.7652

Epoch 1/2 Step 910... Discriminator Loss: 0.9292... Generator Loss: 1.4318
Epoch 1/2 Step 920... Discriminator Loss: 0.9113... Generator Loss: 1.0068
Epoch 1/2 Step 930... Discriminator Loss: 0.9685... Generator Loss: 1.1394
Epoch 2/2 Step 940... Discriminator Loss: 1.2466... Generator Loss: 0.6814
Epoch 2/2 Step 950... Discriminator Loss: 0.8876... Generator Loss: 1.1954
Epoch 2/2 Step 960... Discriminator Loss: 0.9523... Generator Loss: 1.2135
Epoch 2/2 Step 970... Discriminator Loss: 0.9246... Generator Loss: 1.5266
Epoch 2/2 Step 980... Discriminator Loss: 0.9256... Generator Loss: 1.1807
Epoch 2/2 Step 990... Discriminator Loss: 1.8458... Generator Loss: 2.3669
Epoch 2/2 Step 1000... Discriminator Loss: 0.9611... Generator Loss: 1.0115

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 10    # 64
z_dim = 100
learning_rate = 0.0005 # 0.0002
beta1 = 0.5


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch 1/1 Step 10... Discriminator Loss: 2.8314... Generator Loss: 0.1731
Epoch 1/1 Step 20... Discriminator Loss: 0.6363... Generator Loss: 2.2165
Epoch 1/1 Step 30... Discriminator Loss: 0.4024... Generator Loss: 6.5740
Epoch 1/1 Step 40... Discriminator Loss: 0.7629... Generator Loss: 1.6922
Epoch 1/1 Step 50... Discriminator Loss: 0.5296... Generator Loss: 2.1313
Epoch 1/1 Step 60... Discriminator Loss: 3.2863... Generator Loss: 20.0966
Epoch 1/1 Step 70... Discriminator Loss: 1.0840... Generator Loss: 1.7861
Epoch 1/1 Step 80... Discriminator Loss: 2.5683... Generator Loss: 0.3180
Epoch 1/1 Step 90... Discriminator Loss: 1.4681... Generator Loss: 1.7704
Epoch 1/1 Step 100... Discriminator Loss: 1.3640... Generator Loss: 0.8916

Epoch 1/1 Step 110... Discriminator Loss: 1.2000... Generator Loss: 0.8183
Epoch 1/1 Step 120... Discriminator Loss: 2.3167... Generator Loss: 0.2422
Epoch 1/1 Step 130... Discriminator Loss: 0.8602... Generator Loss: 1.0703
Epoch 1/1 Step 140... Discriminator Loss: 0.8595... Generator Loss: 5.6885
Epoch 1/1 Step 150... Discriminator Loss: 2.4609... Generator Loss: 7.8141
Epoch 1/1 Step 160... Discriminator Loss: 0.4496... Generator Loss: 3.2793
Epoch 1/1 Step 170... Discriminator Loss: 0.5671... Generator Loss: 2.6806
Epoch 1/1 Step 180... Discriminator Loss: 0.5874... Generator Loss: 1.8620
Epoch 1/1 Step 190... Discriminator Loss: 0.8104... Generator Loss: 1.3738
Epoch 1/1 Step 200... Discriminator Loss: 2.4155... Generator Loss: 0.2216

Epoch 1/1 Step 210... Discriminator Loss: 0.7630... Generator Loss: 1.5564
Epoch 1/1 Step 220... Discriminator Loss: 0.8994... Generator Loss: 0.9905
Epoch 1/1 Step 230... Discriminator Loss: 1.2678... Generator Loss: 0.7678
Epoch 1/1 Step 240... Discriminator Loss: 1.9604... Generator Loss: 0.2841
Epoch 1/1 Step 250... Discriminator Loss: 0.7293... Generator Loss: 1.7493
Epoch 1/1 Step 260... Discriminator Loss: 2.1849... Generator Loss: 0.2193
Epoch 1/1 Step 270... Discriminator Loss: 1.0048... Generator Loss: 0.9612
Epoch 1/1 Step 280... Discriminator Loss: 1.4089... Generator Loss: 0.5221
Epoch 1/1 Step 290... Discriminator Loss: 1.4643... Generator Loss: 1.4157
Epoch 1/1 Step 300... Discriminator Loss: 1.0593... Generator Loss: 1.1607

Epoch 1/1 Step 310... Discriminator Loss: 1.2002... Generator Loss: 1.7568
Epoch 1/1 Step 320... Discriminator Loss: 1.3758... Generator Loss: 0.7787
Epoch 1/1 Step 330... Discriminator Loss: 0.6306... Generator Loss: 2.0808
Epoch 1/1 Step 340... Discriminator Loss: 0.6094... Generator Loss: 2.0302
Epoch 1/1 Step 350... Discriminator Loss: 1.1749... Generator Loss: 4.1301
Epoch 1/1 Step 360... Discriminator Loss: 1.0593... Generator Loss: 0.8707
Epoch 1/1 Step 370... Discriminator Loss: 1.1397... Generator Loss: 1.2742
Epoch 1/1 Step 380... Discriminator Loss: 2.6071... Generator Loss: 0.2827
Epoch 1/1 Step 390... Discriminator Loss: 0.8784... Generator Loss: 1.4767
Epoch 1/1 Step 400... Discriminator Loss: 0.9946... Generator Loss: 0.9498

Epoch 1/1 Step 410... Discriminator Loss: 0.6361... Generator Loss: 3.6403
Epoch 1/1 Step 420... Discriminator Loss: 1.3175... Generator Loss: 0.5469
Epoch 1/1 Step 430... Discriminator Loss: 1.0056... Generator Loss: 1.0084
Epoch 1/1 Step 440... Discriminator Loss: 1.1850... Generator Loss: 0.7721
Epoch 1/1 Step 450... Discriminator Loss: 2.2220... Generator Loss: 0.2198
Epoch 1/1 Step 460... Discriminator Loss: 1.1750... Generator Loss: 1.3575
Epoch 1/1 Step 470... Discriminator Loss: 1.7230... Generator Loss: 0.4415
Epoch 1/1 Step 480... Discriminator Loss: 0.8890... Generator Loss: 1.1558
Epoch 1/1 Step 490... Discriminator Loss: 1.4923... Generator Loss: 0.4902
Epoch 1/1 Step 500... Discriminator Loss: 1.1658... Generator Loss: 0.6965

Epoch 1/1 Step 510... Discriminator Loss: 1.1473... Generator Loss: 0.8954
Epoch 1/1 Step 520... Discriminator Loss: 0.9212... Generator Loss: 2.8562
Epoch 1/1 Step 530... Discriminator Loss: 1.1809... Generator Loss: 0.9307
Epoch 1/1 Step 540... Discriminator Loss: 1.2080... Generator Loss: 0.7578
Epoch 1/1 Step 550... Discriminator Loss: 0.9365... Generator Loss: 1.1017
Epoch 1/1 Step 560... Discriminator Loss: 0.8107... Generator Loss: 2.1666
Epoch 1/1 Step 570... Discriminator Loss: 1.1199... Generator Loss: 1.8468
Epoch 1/1 Step 580... Discriminator Loss: 0.8234... Generator Loss: 1.6758
Epoch 1/1 Step 590... Discriminator Loss: 1.0412... Generator Loss: 0.7422
Epoch 1/1 Step 600... Discriminator Loss: 1.0246... Generator Loss: 0.8673

Epoch 1/1 Step 610... Discriminator Loss: 0.8925... Generator Loss: 1.2483
Epoch 1/1 Step 620... Discriminator Loss: 0.6509... Generator Loss: 1.6252
Epoch 1/1 Step 630... Discriminator Loss: 1.0243... Generator Loss: 0.7955
Epoch 1/1 Step 640... Discriminator Loss: 1.6829... Generator Loss: 0.3558
Epoch 1/1 Step 650... Discriminator Loss: 1.1385... Generator Loss: 1.0369
Epoch 1/1 Step 660... Discriminator Loss: 1.3515... Generator Loss: 0.5948
Epoch 1/1 Step 670... Discriminator Loss: 1.6330... Generator Loss: 0.3643
Epoch 1/1 Step 680... Discriminator Loss: 0.5455... Generator Loss: 2.1980
Epoch 1/1 Step 690... Discriminator Loss: 0.8668... Generator Loss: 1.2865
Epoch 1/1 Step 700... Discriminator Loss: 1.1457... Generator Loss: 1.3698

Epoch 1/1 Step 710... Discriminator Loss: 0.6788... Generator Loss: 2.8326
Epoch 1/1 Step 720... Discriminator Loss: 0.9228... Generator Loss: 0.9962
Epoch 1/1 Step 730... Discriminator Loss: 1.4937... Generator Loss: 0.6488
Epoch 1/1 Step 740... Discriminator Loss: 1.2856... Generator Loss: 0.6185
Epoch 1/1 Step 750... Discriminator Loss: 0.9533... Generator Loss: 1.8533
Epoch 1/1 Step 760... Discriminator Loss: 1.2532... Generator Loss: 0.9895
Epoch 1/1 Step 770... Discriminator Loss: 0.7572... Generator Loss: 1.8864
Epoch 1/1 Step 780... Discriminator Loss: 1.0193... Generator Loss: 1.0729
Epoch 1/1 Step 790... Discriminator Loss: 1.0442... Generator Loss: 0.9879
Epoch 1/1 Step 800... Discriminator Loss: 1.0397... Generator Loss: 1.3141

Epoch 1/1 Step 810... Discriminator Loss: 1.3552... Generator Loss: 0.5020
Epoch 1/1 Step 820... Discriminator Loss: 0.6109... Generator Loss: 1.8941
Epoch 1/1 Step 830... Discriminator Loss: 1.3475... Generator Loss: 2.0259
Epoch 1/1 Step 840... Discriminator Loss: 1.2856... Generator Loss: 1.2052
Epoch 1/1 Step 850... Discriminator Loss: 0.9706... Generator Loss: 1.2808
Epoch 1/1 Step 860... Discriminator Loss: 1.4629... Generator Loss: 0.4401
Epoch 1/1 Step 870... Discriminator Loss: 1.3063... Generator Loss: 0.7970
Epoch 1/1 Step 880... Discriminator Loss: 1.1650... Generator Loss: 0.6655
Epoch 1/1 Step 890... Discriminator Loss: 0.9150... Generator Loss: 1.7967
Epoch 1/1 Step 900... Discriminator Loss: 1.1097... Generator Loss: 1.2153

Epoch 1/1 Step 910... Discriminator Loss: 0.9035... Generator Loss: 1.0795
Epoch 1/1 Step 920... Discriminator Loss: 0.9590... Generator Loss: 1.6811
Epoch 1/1 Step 930... Discriminator Loss: 1.4152... Generator Loss: 0.5315
Epoch 1/1 Step 940... Discriminator Loss: 0.5391... Generator Loss: 3.6922
Epoch 1/1 Step 950... Discriminator Loss: 1.1208... Generator Loss: 0.8245
Epoch 1/1 Step 960... Discriminator Loss: 1.1278... Generator Loss: 0.9015
Epoch 1/1 Step 970... Discriminator Loss: 1.3310... Generator Loss: 0.6133
Epoch 1/1 Step 980... Discriminator Loss: 1.7753... Generator Loss: 0.3056
Epoch 1/1 Step 990... Discriminator Loss: 0.7518... Generator Loss: 1.5241
Epoch 1/1 Step 1000... Discriminator Loss: 1.1132... Generator Loss: 0.9499

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.